Markets and Hierarchies

A market once operated inside the factory

At the Winchester Repeating Arms Company in the late nineteenth century, a visitor could have crossed a factory floor without crossing cleanly from "market" into "hierarchy." Winchester owned the building and machines, supplied materials, and sold the finished guns. Yet much of the work was managed by inside contractors. Each contractor agreed with the company to deliver a part or perform an operation, then hired, assigned, supervised, and paid the people who did it. In 1880, even successive operations on a gun barrel could belong to different contractors. ¹

The arrangement gave contractors strong incentives and relieved the company of daily supervision. It also produced a factory whose departments could pursue their own advantage. A contractor controlled wages and work methods while the company tried to coordinate quantities, quality, machinery, and the flow of parts across the whole plant. When the system receded, the workers did not move to a different workplace; many moved from a contractor's authority to the company's direct employment. ²

That history comes from John Buttrick's 1952 study of Winchester, not from fieldwork conducted by Oliver Williamson. Buttrick reconstructed the system from company records to explain its operation and decline. Williamson adopted inside contracting as a revealing organizational failure in Markets and Hierarchies. It gave physical form to the book's central question: when do repeated exchanges need more common administration than a set of prices can provide? ³

Antitrust supplied the carrying problem

Williamson arrived at that question by a route that joined economics, organization theory, and public policy. His training at Carnegie placed bounded rationality and administrative process beside price theory. Work as a special economic assistant in the U.S. Department of Justice's Antitrust Division in 1966–1967 then confronted him with vertical integration: should bringing a supplier or distributor inside a firm be treated mainly as a restriction on competition, or could it solve a coordination problem that ordinary contracting handled badly? ⁴

In a later account of the book's development, Williamson described a 1969 University of Pennsylvania seminar on “Theories of Institutions” as the setting in which he assembled these strands. A 1971 paper preceded the 1975 book. Ronald Coase had already made the cost of using prices central in The Nature of the Firm. Williamson changed the level of resolution. Rather than ask only why a firm exists, he compared particular transactions and the governance structures available to carry them. That documented line of influence does not make the 1975 framework a simple restatement of Coase. ⁵

The argument is built from hazards, not labels

The book begins with two behavioral conditions. People are boundedly rational: they cannot foresee and write a contract for every future state. Some may also behave opportunistically, disclosing information selectively or exploiting a dependence when the chance arises. Uncertainty and complexity therefore matter because adaptation will be needed after the parties have committed themselves. Small-number exchange matters because competition may no longer discipline the relationship once alternatives disappear. Williamson called the resulting asymmetry “information impactedness.” ⁶

Idiosyncratic investment makes this sequence concrete. A supplier may buy a machine, locate a plant, or develop know-how that is worth much more in one relationship than anywhere else. Before that investment, several partners may compete. After it, exit destroys value, and the parties can bargain over a surplus they created together. A long-term contract, common ownership, or an internal administrative process may protect adaptation better than repeated spot exchange. Each safeguard has costs of its own. ⁷

Williamson's comparison is therefore not "firms are better than markets." Markets preserve choice and strong price incentives. Hierarchies can coordinate interdependent changes and settle some disputes through authority, but their administration can distort information and weaken incentives. Intermediate arrangements combine parts of both. Later transaction-cost economics condensed the comparison around asset specificity, uncertainty, and frequency. Those terms illuminate the 1975 argument, but the later standard formulation should not be mistaken for the book's exact original vocabulary. ⁸

Inside contracting shows why the distinction matters. Buttrick's Winchester is a historical case with records, dates, and participants. Williamson's use of it is an interpretation: divided control over successive production stages made local incentives costly to reconcile with the company's need for system-wide adaptation. The case illustrates his mechanism; it does not by itself test every claim about transaction attributes or prove that direct employment was the only possible remedy. ⁹

A hierarchy changes the worker's relation, not just the transaction's cost

Winchester's transition does not furnish a simple cost-saving result. Buttrick reports that replacing contractors with company foremen actually increased costs in virtually every converted job. A superintendent nevertheless judged the jobs better run because the company had better control. The episode is evidence that management preferred consolidated control, not an unambiguous empirical test of Williamson's efficiency explanation. ¹⁰

The transition also redistributed authority. An inside contractor had room to choose methods and could profit from organizing a department, while employees depended on that contractor for hiring and pay. Under direct management, the company acquired more consistent control over both contractors and workers. What appears from the executive office as a more governable arrangement appears from the floor as a change in who gives orders, sets rates, hears a grievance, and captures the gain. ¹¹

The same records show how narrow a transaction-cost account can be. Winchester fired six machinists identified as union leaders and marked people suspected of union activity in personnel records with “robal”—“labor” reversed. These events do not prove that one governance form caused the repression. They do show that worker voice and managerial control were part of the historical institution that Williamson's organizational illustration compresses into an efficiency problem. ¹²

That is where cooperation, incentives, and organizational equilibrium adds something the efficiency comparison cannot supply alone. A transaction may remain inside a hierarchy because exit is costly, not because affected people regard its authority as legitimate. Worker voice, collective bargaining, law, racial and gender exclusion, and control over dangerous conditions do not automatically enter a calculation built from the contracting parties' private costs. ¹³

The same boundary problem reaches beyond employment. Vertical integration can protect a specialized supplier while concentrating market power. Outsourcing can lower a focal firm's costs by moving injury, pollution, or care obligations onto people and ecosystems outside its accounting boundary. A governance arrangement can economize on bargaining between two parties while making other beings bear the adaptation. ¹⁴

The framework survived by becoming disputable

Mark Granovetter's 1985 essay on embeddedness directly challenged the markets-and-hierarchies research program. He argued that economic action in both forms is carried by continuing social relations. Trust can support exchange without integration, while hierarchy does not abolish malfeasance; authority itself is embedded in networks, histories, and obligations. This was a theoretical dispute, not a statistical test, and it changed the question from which formal structure governs a hazard to which actual relationships make a structure work. ¹⁵

Empirical research produced a more variegated verdict. Robert David and Shin-Kap Han's 2004 systematic assessment examined 308 statistical tests reported in 63 articles. It found comparatively stronger support around asset specificity, considerable disagreement about how central constructs should be measured, and lower support around uncertainty and performance. That review tests a research program developed after 1975, not Buttrick's case or every proposition in the book. Its mixed result is useful precisely because “transaction cost” otherwise risks becoming an explanation attached after any boundary choice. ¹⁶

The durable practice is a sequence of inquiry. Name the adaptation or bargaining problem; identify whose investment creates dependence; compare the concrete safeguards and the failures each introduces; then widen the accounting to the people and places whose costs the transaction does not record. Used that way, Markets and Hierarchies makes organizational boundaries investigable without making efficiency the final word on how authority should be arranged. ¹⁷

The links to bounded rationality, structure and scale, governance and accountability, cooperation and equilibrium, organizational intelligence, and benefit for all life are editorial comparisons of cognition, authority, participation, and consequence. Only the relation to Coase is presented as a documented intellectual influence; the other links do not assert direct influence or equivalent conclusions. ¹⁸